The goal of this TUT capital gain funded project is to study new methods and applications of big data. The project is multi-disciplinary and consists of six different topics. Our topic is to investigate potential of big visual data and efficient methods for processing a huge amount of visual information (images and videos). In particular, we want to understand the complex networks of millions of images and be able to detect and learn various visual classes and their attributes unsupervised or semi-unsupervised from data and metadata.


Fatemeh Shokrollahi E-mail Research Student
Jukka Lankinen E-mail Research Student
Dr. Ke (Cory) Chen E-mail Post-doc Researcher
Joni Kamarainen E-mail Professor

Selected Publications

Local Feature Based Unsupervised Alignment of Object Class Images

By J. Lankinen and J.-K. Kämäräinen In British Machine Vision Conference (BMVC) 2011.

Local feature based image alignment to a selected seed image.

Download: PDF, BibTex, Source Code

Discovering Multi-Relational Latent Attributes by Visual Similarity Networks

By F. Shokrollahi~Yancheshmeh and J.-K. Kämäräinen and K. Chen In Asian Conf. on Computer Vision (ACCV) Workshop on Feature and Similarity 2014.

Our method automatically constructs a visual similarity graph and finds "multi-view" graph paths between any two images.

Download: PDF, BibTex, Source Code

Unsupervised Visual Alignment with Similarity Graphs

By F. Shokrollahi~Yancheshmeh, K. Chen and J.-K. Kämäräinen In IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2015.

Our method constructs a visual similarity network, selects a seed image and then aligns other images to the seed by step-wise graph travelling.

Download: PDF, BibTex, Source Code


Data sets

"Baseline" set
Pascal VOC
The annual Pascal VOC challenge
Thousands of labelled images in WordNet hierarchy
Millions of labelled images in WordNet hierarchy (the preferred dataset)

Authors and teams

Pedro Felzenszwalb
The Part-Based Method for Object Detection and Classification
Fei-Fei Li
The mother of Caltech-101 and later ImageNet
Andrew Zisserman's group
Bag of Visual Bag of Words methods starting 2003 (typically heavily SVM oriented)
Erik G. Miller
Image congealing works (related to unsupervised object alignment)